Image('images/pydata_logo.png', width='300px')
IFrame('https://pydata.org/code-of-conduct/', **default_size)
IFrame('https://numfocus.org/sponsored-projects', **default_size)
Image('images/pydata-map.png', **default_size)
Image('images/pydata-uk-map.png', **default_size)
Image('images/pydata-london-screenshot.png', **default_size)
Come join the community on on Slack, Twitter or Linkedin
Come join the community on our UK-wide Slack channel!
Image('images/slack_join_qr.png', width='360px')
Image('images/twitter_qr_code.png', width='360px')
Image('images/linkedin_qr_code.png', width='360px')
Image('images/pretalx_cfp.png', width='600px')
Image('images/pydata_london_cfp.png', width='360px')
Image('images/pydata_london_2023.png', **default_size)
This talk is about my journey of building a public-facing Wordle Solver meant to be future-proof. It showcases the possibility of rapidly developing a low-maintenance and public-facing side project with the help of unsupervised learning.
The Wordle game captivates human players and people who want to build machine players. The developers of most Wordle solvers strive to find the best or optimal solution to the game by using supervised learning targeting a specific word list. However, these approaches may take a long time to train a model and easily overfit / only work well with a specific Wordle implementation.
Despite time constraints, I built a Wordle Solver that works with multiple Wordle versions and supports variable word lengths. Moreover, the World Solver is now running with almost zero maintenance and at a low operating cost.